What is the MAX number of CPU cores/threads that Acronis can use during a backup?
I have an AMD 3900x CPU with 12 cores/24 threads.
On my previous system (Intel 7700k 4 cores/8 threads) the CPU would be at 90% during a backup.
It was backing up from a Samsung 960 EVO 1TB SSD to a Western Digital WD1003FZEX 1TB drive, the backup would take approx 33 minutes.
Now on the new hardware, the CPU use is approx 25% (5-6 threads) using MAX compression for the same data.
Backup from a new Corsair Force MP600 1TB SSD (PCIe 4) to a Western Digital WD4005FZBCX 4TB drive takes approx 6 1/2 mins, if I back up to the Samsung 1TB SSD it takes 4 1/2 minutes.
I know I am being picky but it would be great if Acronis used more cores (if possible) to cut the backup time even more.
The data amount being backup up is approx 113GB, backed up size is approx 61GB
Hi,
I already changed it from Low to Normal and it did not make any difference.
I changed it ti High and ran it, it appeared to average approx 25% CPU use, so no change.
I know it only runs for approx 3 1/2 mins instead of 35 mins previously so it is not a problem.
Just would like to know if Acronis have set limits on how many CPU cores it will use.
To the best of my knowledge there is no limit on number of CPU cores. An easy way to find out is open Windows Task Manager then, click on the Performance tab. At the bottom of the window look for Open Resource Monitor and click on that.
This will open the Resource Monitor where you can see details on all system resources. Click on the CPU tab. In the right side column you will see CPU Total followed by Service CPU Usage followed all cores starting with CPU 0.
Now run the backup and have a look at each core.
Check the CPU affinity on trueimage.exe to see if it shows all cores in use by default, or not.
Checked that, ALL CPU cores are set to be used by default.
Yes, that seems to be the default (all CPUs available, including hyperthreaded cores)- on Ryzen 7 1700X system all 16 are ticked. Have not noticed that backups are noticeably quicker on it compared to my i5-9400 system where there are only 6.
Ian
There you go!
Technically it is capable of using all of them then.
Whether or not it does though, that may be something only the developers can tell us. It really depends on how the processes are programmed to run and take advantage of available cores.
The best you can probably do is monitor it while in use with something like CPU-Z. You can download it for free (there is a portable version so you don't actually have to install it if you don't want to). Launch it and monitor CPU usage before a backup, then start the backup and see how the cores ramp up (or not).
Or use Taskmgr - Performance View - CPU view and shows all cores
Hello Martin,
I have a few comments from my experience with Acronis True Image and system performance in general. These are all inter-related, so there is no one "silver bullet". When you improve efficiency in one area, you then become limited by the next bottleneck that was masked by the first issue. Sometimes, you can improve efficiency ... but overall performance does not improve because the issue was not the most critical bottleneck.
I run a 8-core Xeon workstation, 48 GB of memory, with 4 SSDs (of varying speeds) with 7 hard disks (of varying speeds). I find that Acronis True Image 2016 runs well, when monitored by Microsoft Sysinternals Process Explorer.
I highly recommend that you try Process Explorer. It is the most flexible Windows performance monitor I have found, and it is free.
A few basics.
Overall, your performance is a combination of how fast you can read the source data, process it (check for incremental/differential mode, compress) and the speed that you can write to the target.
Acronis will overlap the source scanning, compressing and output, but ultimately, you can't go faster than your target.
You may need to try some experiments on your configuration to see if there is a net gain.
Also realize the amount of work done by Acronis varies substantially if you are performing a "full", "Differential" or "Incremental" mode. For Differential and Incremental mode, the block change tracking feature makes a big difference .. if it is being used.
Hare are a few of the big points.
Run the process explorer monitor, set the update time to 10 seconds, and open the chart window. The charts will show the total data read and written per sample (which is 10 seconds). Divide by 10 and you know how fast overall the source disk and target disks are being read/written.
Source disk activity is more random. On SSDs it is important that the partitions are aligned on 1 MB boundaries or larger. This happens by default on Windows 10 when it partitions a new disk. Disks formatted in older versions, or USB disks formatted by other companies may not be properly aligned, so check.
Target disk IO is primarily sequential large block reads (to compare against the previous backup) and writes.
The default Windows NTFS cluster size of 4KiB is NOT optimized for large IO and is NOT efficient.
If your target disk is used primarily as a backup target, I would recommend reformatting the disk with a 64 KiB cluster size. This reduces the metadata overhead of writing large files 16-fold compared to the 4 KiB cluster. Microsoft recommends using 64 KiB cluster size for any file system handling large files. It is not space-efficient when handling small files, but we don't have that case.
Don't use USB2 disks ... or plug a USB3 disk into a USB2 port. USB2 is realistically limited to about 25 MiB/sec max. This shows up as 250 MiB in the Process Explorer 10-second sample.
Running in large-block sequential mode, ...WITH A 64 KiB cluster size, most disks can run faster than 25 MiB/sec.
BTW ... disk fragmentation on the target does make a difference. For good performance, freshly format the target with a 64 KiB cluster ... and the backup images will be nearly contiguous. You don't need to defrag the target disk in most cases.
USB2 disk ... 4KiB cluster ... slowest
Any SATA or USB3 ... the pipe is faster than the disk.
How fast is the disk? Some disks are only 5400 RPM and may only reach 40-50 MB/sec sequential. 7200 RPM disks are faster and run at 65 MB/sec worst case, and 90+ MB/sec on the faster/denser disks ... like the HGST Helium series. They also cost more. My 7200 RPM 4 TB and 8 TB HGST disks run over 100 MB/sec as Acronis targets (using 64 KiB clusters).
CPU ..Hyperthreads ... it depends. Each hyperthread only runs about 75% of the speed of the core, so if you are NOT using all the hyperthreads, you may be running slower. Disabling hyperthreads allows each full-thread to now run at full speed. On CPUs with 4 or less cores, hyperthreads is probably a net benefit for personal use. More than 4 cores, you may be better off disabling hyperthreads. Each "compute-bound" thread gets more throughput. For virtualized server workloads, hyperthreads is normally beneficial.
When I am backing up NVME SSD source to 100 GB/sec 7200 RPM 4+ TB target disks with 64 KiB clusters, I run over 100 MiB/sec on the target disk. That is fast enough for me.
Someone mentioned the Acronis CPU Priority setting. Yes, you should set it to maximum for the fastest backup.
Extra-optimization ... SSDs are fast and handle random IO well ... but are faster still if they can do larger sized IO because their files are contiguous. So improving contiguous-ness can help ... but you do NOT want to "defrag" a SSD ... it generates too much write activity and will impact the useful life.
You can open a command window in administrator mode, and run "defrag /a c:" to analyze your boot SSD disk to see how fragmented it may be. I also un-check my target backup disks from automatic defrag (from the GUI defrag app), and use the command line to report on the fragmentation when I need to check.
In my case, I used Acronis tools to clone the SSD to a hard disk temporarily. It can be an inexpensive USB3 disk.
Then I defrag'ed the hard disk clone version. You can get reports on the disk on how badly it was defragmented .. and even after a fresh Win10 install, the resulting disk is fairly fragmented.
The hard disk defrag will take some time ... a few hours is typical.
Once the defrag is done, then I copy back the now-defragged disk image back onto the SSD ... and voila, I have a defragged SSD, that only took one write per block to create.
You have to use the bootable disk cloning/defrag tools to get this one ... and you need to be careful not to screw up the boot order.
This may be overkill ...
Also .... When the Acronis backup starts, a volume shadow snapshot is issued. This can take some time. You can see this in Process Explorer as the "vss_xxx" processes consuming time.
Acronis Customer,
Thank you for the comprehensive write up on the subject of performance. I am in complete agreement with you and practice backup and data handling much the same as you.
I too use the command "dfrag /a driveletter: command on my computers with some regularity and find it most useful. Having not done that in some time I figured I should do it so I did and captured the results. I am posting them here so as to encourage others to do the same and see what results they have. :)
Microsoft Windows [Version 10.0.18362.295]
(c) 2019 Microsoft Corporation. All rights reserved.
C:\WINDOWS\system32>defrag /a c:
Microsoft Drive Optimizer
Copyright (c) Microsoft Corp.
Invoking analysis on Windows 10 PRO (C:)...
The operation completed successfully.
Post Defragmentation Report:
Volume Information:
Volume size = 475.76 GB
Free space = 341.64 GB
Total fragmented space = 7%
Largest free space size = 186.23 GB
Note: File fragments larger than 64MB are not included in the fragmentation statistics.
You do not need to defragment this volume.
C:\WINDOWS\system32>defrag /a D:
Microsoft Drive Optimizer
Copyright (c) Microsoft Corp.
Invoking analysis on Old Users Profiles (D:)...
The operation completed successfully.
Post Defragmentation Report:
Volume Information:
Volume size = 931.38 GB
Free space = 778.94 GB
Total fragmented space = 0%
Largest free space size = 733.81 GB
Note: File fragments larger than 64MB are not included in the fragmentation statistics.
You do not need to defragment this volume.
C:\WINDOWS\system32>defrag /a E:
Microsoft Drive Optimizer
Copyright (c) Microsoft Corp.
Invoking analysis on VSS Disk (E:)...
The operation completed successfully.
Post Defragmentation Report:
Volume Information:
Volume size = 931.49 GB
Free space = 884.67 GB
Total fragmented space = 0%
Largest free space size = 747.63 GB
Note: File fragments larger than 64MB are not included in the fragmentation statistics.
You do not need to defragment this volume.
C:\WINDOWS\system32>defrag /a F:
Microsoft Drive Optimizer
Copyright (c) Microsoft Corp.
Invoking analysis on App & User Data (F:)...
The operation completed successfully.
Post Defragmentation Report:
Volume Information:
Volume size = 465.63 GB
Free space = 342.09 GB
Total fragmented space = 0%
Largest free space size = 338.63 GB
Note: File fragments larger than 64MB are not included in the fragmentation statistics.
You do not need to defragment this volume.
C:\WINDOWS\system32>defrag /a G:
Microsoft Drive Optimizer
Copyright (c) Microsoft Corp.
Invoking analysis on Backup & Mass Storage (G:)...
The operation completed successfully.
Post Defragmentation Report:
Volume Information:
Volume size = 1.81 TB
Free space = 1.33 TB
Total fragmented space = 0%
Largest free space size = 545.49 GB
Note: File fragments larger than 64MB are not included in the fragmentation statistics.
You do not need to defragment this volume.
C:\WINDOWS\system32>defrag /a I:
Microsoft Drive Optimizer
Copyright (c) Microsoft Corp.
Invoking analysis on Quick BU-Scratch Disk (I:)...
The operation completed successfully.
Post Defragmentation Report:
Volume Information:
Volume size = 232.86 GB
Free space = 2.68 MB
Total fragmented space = 0%
Largest free space size = 6.25 MB
Note: File fragments larger than 64MB are not included in the fragmentation statistics.
You do not need to defragment this volume.
C:\WINDOWS\system32>
As mentioned earlier, I am doing a backup (FULL each day) from my Corsair MP600 1TB PCIe4 SSD to my Samsung 850 EVO Proc PCIe3 SSD.
My previous system was an Intel 7700K CPU backing up (FULL) my Samsung 850 EVO SSD to a Western Digital Black 1TB drive.
New system is AMD 3900x CPU doing the backup from Corsair to Samsung SSD's.
Backup time on new system was approx 4 1/2 minutes Input size approx 77GB Output size approx 35GB - 37GB
I formatted the Samsung SSD to use 16KB blocks and it took an extra 25 seconds.
I then formatted it to use 32KB blocks and the backup took 3 mins 8 seconds
The first 60 seconds of each backup is preparing before doing the actual backup.
Happy with it but just wish it used more CPU cores as the SSD's have plenty in reserve for writing data out.
Decided to test CPU usage on one of my systems. PC is a Gen 4 Intel i5 4690K @ 3.50GHz. Backed up OS C: drive, Samsung 970 PRO 512GB NVMe PCIe Gen 3 x4 to an identical Samsung Pro NVMe PCIe drive. C: drive holds 144GB of data. ran a FULL backup with performance priority set to High. Output size was 46.5GB. Drives have a 64KB block size.
Total time to run backup was 5 minutes 47 seconds. CPU load all 4 cores was 111% for most of the backup time, (system is water cooled). See screenshot below:
Immediately after the above task finished I configured a new task and set performance priority to Low. CPU usage remained at 111% as previous run. Total time to run 5 minutes and 57 seconds.
There have been issues in prior versions of TI where Priority did not work properly. This may be the case again.
Ultimately, I think Acronis Customer hit the nail on the head. Disk performance is probably the biggest contributor to performance in backup programs. Doesn't matter how many cores you have if the drive has a max write limitation... And that limitation is really the 4K read/write speeds.
I agree with Rob in that the largest contributor to performance is disk read/write limitation. If you have fast drives this will dictate performance more than any other component.